Objekterkennung und -verortung in Luftbildaufnahmen für Such- und Rettungseinsätze mittels neuronaler Faltungsnetzwerke

The bachelor thesis ‚Object detection and localization in aerial photographs for search and rescue tasks with convolutional neural networks’ is concerned with the detection and localization of essential objects in aerial photographs from disaster areas taken by unmanned aerial vehicles. For the detection of objects in aerial photography the approach artificial neural networks (ANN) from the field of machine learning is presented and a convolution neuronal network is trained by means of a library. In addition, an approach for the localization of the detected objects is presented. This type of information retrieval is intended to provide rescue workers with a more accurate indication of the extent of the disaster and to assist in the planning, coordination and execution of further rescue operations. The work concludes with a detailed evaluation using real aerial photographs